Hi, researchers,
recently I refined a structure using FullProf, the result seems weird.
The Chi2 is much close to 1, namely 1.33, while the Rwp is still high as 14.6, as the attached Fig1. shows.
The same data I have refined using GSAS seems have a different result, the wRp in GSAS is like normal, as the Fig2. shows.
What causes this difference and how can I improve my refinement?
Something is wrong, could you post the GSAS liveplot.
The Repx is so different (Fullprof 12.7, GSAS 3.4).
Are you sure refine the same data with the same CIF?
Try to refinement with the same parameters for both.
Dear Xu Mingquan,
The weighted profile factor Rwp shouldn't be taken alone as a point of comparison, because its value changes significantly with the type of experiment (X-ray/neutron etc), the collection time, and as in your case, even the way refinement is carried out. Rwp is sensitive to the background - the value will change drastically if you do refinement before and after background correction. (Rwp would be lower in the former case).
For that reason, the chi-square factor is better for comparison, since it normalises Rwp with an expected (ideal?) factor (Rexp).
In your case, the chi-square is similar in both software packages, despite different Rwp (and Rexp) values. Hence it is likely that the background is subtracted in one package and not in the other, please check that. If this is not the case, possibly the constraints are different in both. Looking into that can be involved but informative, if that is your objective.
Nevertheless, since the chi-square value is appropriate (close to 1) in both refinements and the fit visually seems quite good, I don't think there is cause to worry.
These papers are highly recommended reading:
https://www.researchgate.net/publication/200045616_Reitveld_Refinement_Guidelines
https://www.researchgate.net/publication/253295652_R_Factors_in_Rietveld_Analysis_How_Good_is_Good_Enough
As for software-package specific differences, I hope someone deeply familiar with GSAS and FullProf will be able to direct you better.
Article Reitveld Refinement Guidelines
Article R Factors in Rietveld Analysis: How Good is Good Enough
At the first screenshot you see the 'conventional' R-factors from Fullprof, which are calculated after applying the background correction (see Fullprof manual, page 44). Your background is quite high, so the corrected R-factors become much higher. GSAS uses different equations for the background-corrected R-factors (GSAS manual, page 174), and the difference between the corrected and uncorrected values is much less pronounced. If you look 2 lines higher in the Fullprof output, you should see the uncorrected R-factors. They are probably much closer to the ones obtained from GSAS.
To Maykel Manawan,
Thanks for your reply.
Yeah, the data and cif file are the same.
Not all the parameters I refined, in the two software, are the same, such as, In GSAS I refine S/L&H/L, in FullProf I refine Asym1&Asym2. But, that would cause such difference? I confused.
I attached the GSAS liveplot.
To Alexander B. Missyul’,
thanks for your reply.
I will refer to the Pages you mentioned.
Additionally, the uncorrected R-factors you mentioned is the highlight in the attached Fig. ?
Try not to exclude the data in FullProf (start from same 2theta as GSAS). The Rexp and Rwp will reduce.
Dear Xu,
Right, this is what you need. You can see that they are slightly lower than for the GSAS refinement, corresponding to the lower chi2.
By the way, I just mentioned that the GSAS file name is NCM. This means that the high background is due to the X-ray fluorescence from cobalt in your sample. Try to find a possibility to work with another X-ray tube (Mo or Ag, for example) or use more narrow energy window of the detector. This will significantly improve the quality of your data.
To dear Alexander B. Missyul’,
nice reply. I have learned a lot from your answer.
Thanks very much.
To dear Alexander B. Missyul’,
additional question: In the FullProf Manual, the conventional R-factors all have a constant c. does it have the same identification as the normalisation c on Page 43?
c is added in front of the R-factor symbols on page 44 in order to denote the conventional R-factor in contrast to the uncorrected ones. It is part of the symbol, not a coefficient.
Dear Alexander B. Missyul’,
In the case of Xu Mingquan, Would it be correct to report the uncorrected R factors in a paper? Or is it necessary to improve the quality of the diffractogram (new measurements) to reduce the Rwp and report the conventional Rwp value?. I have a similar problem to Xu Mingquan with my Rietveld Analysis in Fullprof for the LiTaO3.
Thanks.
- Badges
- Science method
- Similar topics
- Methodology
- Crystallography
- X-ray Diffraction